Method for synchronizing program sections of a computer program

Abstract

The invention relates to a method for synchronizing program sections of a computer program. The program sections can run in parallel on different processors of a computer. Data transfer between the program sections is synchronized by providing a buffer. Unauthorized access to the buffer is prevented by means of a flag that is set automatically by buffer utilities. The data transfer between individual program sections is thus synchronized by the buffer synchronization class only, which consists of the buffer, internal variables and the buffer utilities, is configured in a very simple manner and can thus be tested in operation an in a relatively simple manner by means of a computer comprising several processors. The data transfer is very efficient as the individual utilities are designed in a simple and short manner and thus require little processing time and as the data can be directly written in or read out of the buffer.

Description

BACKGROUND OF THE INVENTION[0001]1. Field of the Invention[0002]The invention is directed to a method for synchronizing program sections of a computer program whose program sections can run in parallel on different processors of a computer.[0003]2. Description of the Related Art[0004]The programming and synchronization of threads is explained in “Win 32 Multithreaded Programming”, Aaron Cohen and Mike Woodring, O'Reilly & Associates Inc., ISBN 1-56592-296-4, particularly on pages 258 through 269 and 301 through 303. Such threads are also referred to as “program sections” below.[0005]Program sections that generate data for another program section are referred to as “producers” or “writers”, and program sections that process such data are referred to as “users” or “readers”. It is necessary that the individual program sections work asynchronously in order to optimally utilize the computing power of a computer having a plurality of processors in the division of the work between the ind...

AI Technical Summary

Benefits of technology

[0015]The invention is based on the object of providing a method for synchronizing program sections of a computer program that can be employed in a versatile manner, allows a very efficient data transfer between the program sections, and is also fashioned in a simple way.

[0019]According to a preferred embodiment of the invention, the individual buffer sections are arranged lying immediately behind one another in the memory area of the computer, and administrative data such as the flag, the size of the buffer section, or the data type allocated to the respective buffer sections are stored in as memory area that is independent therefrom. This creates a continuous memory area occupied with data, and this can be quickly and simply read out or written with new data. Moreover, this fashioning of the buffer allows that this data can be written with blocks of a specific size and blocks of a different size can be read. As a result, the data transfer rate, which is influenced by the size of the blocks, can be correspondingly adapted to the corresponding demands that may be high or low both at the side of the producer program sections as well as at the side of the read program sections.

[0020]An inventive control module (synchronization buffer class) for buffer memory administration can be programmed and / or offered as a turnkey module with which, on the one hand, large quantities of data can be exchanged between various threads and that, on the other hand, allows very flexible read and write operations. A buffer memory in which data are exchanged between the threads is administered such that a buffer write thread can define the size of the buffer section it needs, whereas a buffer read thread can read data across a plurality of buffer sections.

Problems solved by technology

Such an asynchronous data transfer is highly efficient but very hard to program.

One problem of these computer programs distributed onto a plurality of program sections is that a test in which the program sections run simultaneously on a plurality of processors is difficult to implement since different errors can occur depending on how the processing of the individual program sections overlap in terms of time.

This can result in, e.g., a one-in-a-thousand chance of an unpredictable error occurring, even thought the computer program functions correctly every other time.

Images